A Neural Network pressure control approach for automotive Variable Bleed Solenoid

Abstract

In this paper we describe a novel pressure control approach applied to a Variable Bleed Solenoid (VBS) that resides inside a vehicle automatic transmission for clutch engaging purposes. The pressure control approach is based on a Neural Network Scheme that has been previously trained to learn from the inputs and outputs the nonlinear dynamics experimented on an automatic transmission working on the field. The Neural Network control technique shows a better performance, when compared against a classical method based on a look up pressure table, because avoids VBS output pressure instability especially when the system is subjected to oil temperature fluctuation, input pressure instability, current fluctuation and thermal degradation.